MXFP4 Open on Hugging Face
Parameters
4.0B
Weights (est.)
≈ 2.3 GB
Mac memory (est.)
8 GB+
Size is estimated from the parameter count and quantisation. Memory allows a fifth extra for context plus 2 GB for the system; long contexts need more.
Run it
With mlx-lm on an Apple silicon Mac:
pip install mlx-lmmlx_lm.generate --model mlx-community/Qwen3-4B-Instruct-2507-mxfp4 --prompt "Hello"
Details
- Base model
- Qwen/Qwen3-4B-Instruct-2507 (quantized)
- Vendor
- Qwen
- Task
- Text generation
- Quantisation
- MXFP4
- Licence
- apache-2.0
- Downloads
- 82 in the last 30 days · 728 all time · 0 likes
- Published
- 2026-02-08 · updated 2026-02-08 (9mo ago)
- Tags
- qwen3
Other MLX builds of Qwen/Qwen3-4B-Instruct-2507
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Qwen3-4B-Instruct-2507-DDWQ | 4-bit | ≈ 570 MB | 8 GB+ | 38 |
| Qwen3-4B-Instruct-2507-4bit | 4-bit | ≈ 2.3 GB | 8 GB+ | 95K |
| Qwen3-4B-Instruct-2507-4bit-DWQ-2510 | 4-bit DWQ | ≈ 2.3 GB | 8 GB+ | 1.1K |
| Qwen3-4B-Instruct-2507-nvfp4 | 4-bit | ≈ 2.3 GB | 8 GB+ | 185 |
| Qwen3-4B-Instruct-2507-4bit-g32 | 4-bit | ≈ 2.3 GB | 8 GB+ | 90 |
| Qwen3-4B-Instruct-2507-mxfp4 (this one) | MXFP4 | ≈ 2.3 GB | 8 GB+ | 82 |
| Qwen3-4B-Instruct-2507-5.0bpw-dynamic | 4-bit | ≈ 2.3 GB | 8 GB+ | 47 |
| Qwen3-4B-Instruct-2507-5bit | 5-bit | ≈ 2.8 GB | 8 GB+ | 164 |
| Qwen3-4B-Instruct-2507-5bit-g32 | 5-bit | ≈ 2.8 GB | 8 GB+ | 71 |
| Qwen3-4B-Instruct-2507-6bit | 6-bit | ≈ 3.3 GB | 8 GB+ | 221 |
| Qwen3-4B-Instruct-2507-8bit | 8-bit | ≈ 4.3 GB | 8 GB+ | 821 |
| Qwen3-4B-Instruct-2507-mxfp8 | MXFP8 | ≈ 4.3 GB | 8 GB+ | 168 |
| Qwen3-4B-Instruct-2507-bf16 | bf16 | ≈ 8.0 GB | 16 GB+ | 292 |